relu-layer — OpenAI case

OpenAI-inspired neural-network experiment needs a simple nonlinear activation between two learned transformations. You need to apply ReLU elementwise so the team can verify the forward-pass behavior before moving to a larger architecture.

Replace every negative entry with $00$ and keep the others: $\max (0,x)\backslash max(0,x)$ element-wise.

Implement solve(x).

Returns. Return a float NumPy array of the same shape.

Examples

Example 1

Input:

solve([-2, 0, 3])

Output:

[0.0, 0.0, 3.0]

Example 2

Input:

solve([[1.5, -0.5], [-3.0, 4.0]])

Output:

[[1.5, 0.0], [0.0, 4.0]]